The ability to detect campylobacter presence and concentration using different chicken carcass samples

This study aimed to investigate the potential impact of using different sampling types, when testing broiler carcasses for Campylobacter. In Denmark, the current sampling type for surveillance is leg skin (LS) samples and since the EU process hygiene criterion uses pooled neck skin (NS) samples, it is relevant to understand the impact on surveillance outcomes the two sample types have. Neck and skin leg samples were collected from 340 batches slaughtered at the two major broiler slaughter-houses in Denmark. A subset of 156 carcasses were allocated to generate 52 groups and five samples types were derived from each group: 1) single LS; 2) single NS, 3) mean LS, 4) mean NS and e) pooled NS (3 samples). All samples were enumerated for Campylobacter and were considered as positive if >= 10 colony forming units/gram (cfu/g). Prevalence, concentrations, agreement, association and epidemiological sensitivity, as the ability of the sample type to correctly classify a positive group, were compared between sample types. The variability when comparing a single LS to the pooled NS was explored using Monte Carlo analysis. When comparing results on single LS and single NS from individual carcasses, the prevalence was 29.1% (CI95% 24.3%; 34.0%) and 34.7% (29.6%; 38.9%); respectively. The mean concentration of Campylobacter was 199 and 945 cfu/g; respectively. The cfu/g between the two sample types well correlated (R-adj(2) = 0.631). However, the proportion of carcasses with > 1000 cfu/g was significantly higher on NS (15.0%) than on LS (9.4%). At group level, the prevalence derived from single LS samples was lower than all other sample types. Generally, the cfu/g on NS were higher than those on LS. For pooled NS the epidemiological sensitivity and prevalence were approximate to 1.6 times higher than in single LS, but still had moderate agreement (Kappa = 0.56, p < 0.001). The linear relationships between the cfu/g of different sample types was generally good (R-2 between 0.59 and 0.84). In the Monte Carlo analysis, association between the two sample types was found in 91.3% of the scenarios, suggesting some uncertainty in the ability to translate all outcomes between the two sampling types. The study increases understanding of the impact of selecting a sample type for surveillance of Campylobacter in broilers. It could be expected that, by changing from single LS (current system) to the pooled NS, the estimated sensitivity and prevalence could both increase of approximate to 1.6 times.